diff --git a/ratis-grpc/src/main/java/org/apache/ratis/grpc/server/GrpcLogAppender.java b/ratis-grpc/src/main/java/org/apache/ratis/grpc/server/GrpcLogAppender.java
index d1cdb2bb7d..51b13349d6 100644
--- a/ratis-grpc/src/main/java/org/apache/ratis/grpc/server/GrpcLogAppender.java
+++ b/ratis-grpc/src/main/java/org/apache/ratis/grpc/server/GrpcLogAppender.java
@@ -63,6 +63,7 @@
 import java.util.concurrent.ConcurrentHashMap;
 import java.util.concurrent.ExecutionException;
 import java.util.concurrent.TimeUnit;
+import java.util.concurrent.atomic.AtomicBoolean;
 import java.util.concurrent.atomic.AtomicLong;
 import java.util.function.Consumer;
 
@@ -153,6 +154,8 @@ synchronized int process(Event event) {
 
   private final AtomicLong callId = new AtomicLong();
 
+  private final AtomicBoolean snapshotInstallInProgress = new AtomicBoolean();
+
   private final RequestMap pendingRequests = new RequestMap();
   private final int maxPendingRequestsNum;
   private final boolean installSnapshotEnabled;
@@ -275,6 +278,14 @@ private void resetClient(AppendEntriesRequest request, Event event, Throwable er
     }
   }
 
+  private void setSnapshotInstallInProgress(boolean inProgress) {
+    snapshotInstallInProgress.set(inProgress);
+  }
+
+  private boolean isSnapshotInstallInProgress() {
+    return snapshotInstallInProgress.get();
+  }
+
   private boolean isFollowerCommitBehindLastCommitIndex() {
     return getRaftLog().getLastCommittedIndex() > getFollower().getCommitIndex();
   }
@@ -301,8 +312,10 @@ private boolean installSnapshot() {
   public void run() throws IOException {
     try {
       for(; isRunning(); mayWait()) {
+        // Skip sending the snapshot notification and the heartbeat until the heartbeat interval has elapsed.
+        final boolean skipSending = isSnapshotInstallInProgress() && getHeartbeatWaitTimeMs() > 0;
         //HB period is expired OR we have messages OR follower is behind with commit index
-        if (shouldSendAppendEntries() || isFollowerCommitBehindLastCommitIndex()) {
+        if (!skipSending && (shouldSendAppendEntries() || isFollowerCommitBehindLastCommitIndex())) {
           final boolean installingSnapshot = installSnapshot();
           appendLog(installingSnapshot || haveTooManyPendingRequests());
         }
@@ -317,8 +330,10 @@ public void run() throws IOException {
   }
 
   public long getWaitTimeMs() {
-    if (haveTooManyPendingRequests()) {
-      return getHeartbeatWaitTimeMs(); // Should wait for a short time
+    if (haveTooManyPendingRequests() || isSnapshotInstallInProgress()) {
+      // Either the follower has too many pending requests, or it is installing a snapshot
+      // (replies are always IN_PROGRESS/INCONSISTENCY): wait until the next heartbeat.
+      return getHeartbeatWaitTimeMs();
     } else if (shouldSendAppendEntries() && !isSlowFollower()) {
       // For normal nodes, new entries should be sent ASAP
       // however for slow followers (especially when the follower is down),
@@ -749,6 +764,8 @@ public void onNext(InstallSnapshotReplyProto reply) {
       // update the last rpc time
       getFollower().updateLastRpcResponseTime();
       replyState.process(Event.SNAPSHOT_REPLY);
+      // Throttle only while the follower is still installing, and any other reply ends it.
+      setSnapshotInstallInProgress(reply.getResult() == InstallSnapshotResult.IN_PROGRESS);
 
       final long followerSnapshotIndex;
       switch (reply.getResult()) {
@@ -812,6 +829,8 @@ private void handleFailureReply(InstallSnapshotReplyProto reply, BatchLogKey log
 
     @Override
     public void onError(Throwable t) {
+      // The progress of a snapshot notification is unknown after an error, so stop skipping the sending.
+      setSnapshotInstallInProgress(false);
       if (!isRunning()) {
         LOG.info("{} is stopped", GrpcLogAppender.this);
         return;
diff --git a/ratis-test/src/test/java/org/apache/ratis/grpc/TestInstallSnapshotNotificationRpcCount.java b/ratis-test/src/test/java/org/apache/ratis/grpc/TestInstallSnapshotNotificationRpcCount.java
new file mode 100644
index 0000000000..0885d40943
--- /dev/null
+++ b/ratis-test/src/test/java/org/apache/ratis/grpc/TestInstallSnapshotNotificationRpcCount.java
@@ -0,0 +1,231 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package org.apache.ratis.grpc;
+
+import org.apache.ratis.BaseTest;
+import org.apache.ratis.RaftTestUtil;
+import org.apache.ratis.client.RaftClient;
+import org.apache.ratis.conf.RaftProperties;
+import org.apache.ratis.grpc.server.GrpcServicesImpl;
+import org.apache.ratis.proto.RaftProtos.AppendEntriesRequestProto;
+import org.apache.ratis.proto.RaftProtos.InstallSnapshotRequestProto;
+import org.apache.ratis.proto.RaftProtos.RoleInfoProto;
+import org.apache.ratis.protocol.RaftPeerId;
+import org.apache.ratis.server.RaftServer;
+import org.apache.ratis.server.RaftServerConfigKeys;
+import org.apache.ratis.server.impl.MiniRaftCluster;
+import org.apache.ratis.server.impl.RaftServerTestUtil;
+import org.apache.ratis.server.protocol.TermIndex;
+import org.apache.ratis.statemachine.SnapshotInfo;
+import org.apache.ratis.statemachine.StateMachine;
+import org.apache.ratis.statemachine.impl.SimpleStateMachine4Testing;
+import org.apache.ratis.statemachine.impl.SingleFileSnapshotInfo;
+import org.apache.ratis.util.CodeInjectionForTesting;
+import org.apache.ratis.util.JavaUtils;
+import org.apache.ratis.util.SizeInBytes;
+import org.apache.ratis.util.TimeDuration;
+import org.junit.jupiter.api.Assertions;
+import org.junit.jupiter.api.Test;
+import org.junit.jupiter.api.Timeout;
+
+import java.io.File;
+import java.nio.file.Files;
+import java.nio.file.Path;
+import java.util.concurrent.CompletableFuture;
+import java.util.concurrent.TimeUnit;
+import java.util.concurrent.atomic.AtomicBoolean;
+import java.util.concurrent.atomic.AtomicLong;
+import java.util.concurrent.atomic.AtomicReference;
+
+/**
+ * While a follower is installing a snapshot through its state machine (installSnapshotEnabled = false),
+ * the leader should send at most about one notification and one heartbeat per heartbeat interval,
+ * even if a client keeps writing.
+ */
+public class TestInstallSnapshotNotificationRpcCount extends BaseTest {
+  /** The same as RaftServerImpl.INSTALL_SNAPSHOT, which is package-private. */
+  private static final String INSTALL_SNAPSHOT_INJECTION = "RaftServerImpl.installSnapshot";
+
+  private static final int SNAPSHOT_TRIGGER_THRESHOLD = 64;
+  private static final TimeDuration INSTALL_DELAY = TimeDuration.valueOf(3, TimeUnit.SECONDS);
+
+  /** The leader snapshot to install; it is taken by the first notification only. */
+  private static final AtomicReference<SnapshotInfo> LEADER_SNAPSHOT = new AtomicReference<>();
+  private static final AtomicBoolean INSTALLING = new AtomicBoolean();
+  private static final AtomicLong INSTALL_START_NANOS = new AtomicLong();
+  private static final AtomicLong INSTALL_END_NANOS = new AtomicLong();
+
+  // Counted only while INSTALLING.
+  private static final AtomicLong NOTIFICATIONS = new AtomicLong();
+  private static final AtomicLong HEARTBEATS = new AtomicLong();
+  private static final AtomicLong APPEND_ENTRIES_WITH_ENTRIES = new AtomicLong();
+  private static final AtomicLong CLIENT_WRITES = new AtomicLong();
+
+  /** Install the leader snapshot slowly, by copying the snapshot files after {@link #INSTALL_DELAY}. */
+  public static class SlowInstallStateMachine extends SimpleStateMachine4Testing {
+    @Override
+    public CompletableFuture<TermIndex> notifyInstallSnapshotFromLeader(RoleInfoProto roleInfo, TermIndex first) {
+      // With client writes, the leader keeps purging its log so that the follower may be notified again
+      // after the first installation.  Only the first installation is slow and measured.
+      final SingleFileSnapshotInfo leaderSnapshot = (SingleFileSnapshotInfo) LEADER_SNAPSHOT.getAndSet(null);
+      if (leaderSnapshot == null) {
+        return super.notifyInstallSnapshotFromLeader(roleInfo, first);
+      }
+      return CompletableFuture.supplyAsync(() -> {
+        INSTALL_START_NANOS.set(System.nanoTime());
+        INSTALLING.set(true);
+        try {
+          INSTALL_DELAY.sleep();
+          final Path src = leaderSnapshot.getFile().getPath();
+          copy(src);
+          copy(src.resolveSibling(src.getFileName() + ".md5")); // a snapshot without md5 is ignored
+          getStateMachineStorage().loadLatestSnapshot(); // refresh the cached latest snapshot
+          return leaderSnapshot.getTermIndex();
+        } catch (Exception e) {
+          throw new IllegalStateException("Failed to install snapshot", e);
+        } finally {
+          INSTALLING.set(false);
+          INSTALL_END_NANOS.set(System.nanoTime());
+        }
+      });
+    }
+
+    private void copy(Path src) throws Exception {
+      final File dst = new File(getStateMachineDir(), src.getFileName().toString());
+      if (src.toFile().exists() && !dst.exists()) {
+        Files.copy(src, dst.toPath());
+      }
+    }
+  }
+
+  private static RaftProperties newProperties() {
+    final RaftProperties p = new RaftProperties();
+    p.setClass(MiniRaftCluster.STATEMACHINE_CLASS_KEY, SlowInstallStateMachine.class, StateMachine.class);
+    RaftServerConfigKeys.Log.Appender.setInstallSnapshotEnabled(p, false);
+    RaftServerConfigKeys.Snapshot.setAutoTriggerThreshold(p, SNAPSHOT_TRIGGER_THRESHOLD);
+    RaftServerConfigKeys.Snapshot.setAutoTriggerEnabled(p, true);
+    RaftServerConfigKeys.Log.setPurgeGap(p, 8);
+    // Purge the log even though a follower is down, so that it has to install a snapshot.
+    RaftServerConfigKeys.Log.setPurgeUptoSnapshotIndex(p, true);
+    RaftServerConfigKeys.Log.setSegmentSizeMax(p, SizeInBytes.valueOf(1024));
+    return p;
+  }
+
+  @Test
+  @Timeout(value = 60, unit = TimeUnit.SECONDS)
+  public void testRpcCountDuringInstallSnapshotNotification() throws Exception {
+    LEADER_SNAPSHOT.set(null);
+    INSTALLING.set(false);
+    INSTALL_END_NANOS.set(0);
+    for (AtomicLong c : new AtomicLong[] {NOTIFICATIONS, HEARTBEATS, APPEND_ENTRIES_WITH_ENTRIES, CLIENT_WRITES}) {
+      c.set(0);
+    }
+
+    try (MiniRaftClusterWithGrpc cluster = new MiniRaftClusterWithGrpc(
+        MiniRaftCluster.generateIds(3, 0), newProperties(), null)) {
+      cluster.start();
+      final RaftServer.Division leader = RaftTestUtil.waitForLeader(cluster);
+      writeMessages(cluster, leader, SNAPSHOT_TRIGGER_THRESHOLD * 2);
+
+      // Kill a follower and write more, so that the leader purges the log entries the follower needs.
+      final RaftPeerId followerId = cluster.getFollowers().get(0).getId();
+      final long followerNextIndex = cluster.getDivision(followerId).getRaftLog().getNextIndex();
+      cluster.killServer(followerId);
+      writeMessages(cluster, leader, SNAPSHOT_TRIGGER_THRESHOLD * 2);
+      JavaUtils.attempt(() -> Assertions.assertTrue(leader.getRaftLog().getStartIndex() > followerNextIndex),
+          20, ONE_SECOND, "leader purged", LOG);
+      final SnapshotInfo leaderSnapshot = leader.getStateMachine().getLatestSnapshot();
+      LEADER_SNAPSHOT.set(leaderSnapshot);
+
+      CodeInjectionForTesting.put(GrpcServicesImpl.GRPC_SEND_SERVER_REQUEST, (local, remote, args) -> {
+        if (INSTALLING.get() && leader.getId().equals(local) && args[0] instanceof AppendEntriesRequestProto) {
+          final AppendEntriesRequestProto r = (AppendEntriesRequestProto) args[0];
+          if (followerId.toByteString().equals(r.getServerRequest().getReplyId())) {
+            (r.getEntriesCount() == 0 ? HEARTBEATS : APPEND_ENTRIES_WITH_ENTRIES).incrementAndGet();
+          }
+        }
+        return false;
+      });
+      CodeInjectionForTesting.put(INSTALL_SNAPSHOT_INJECTION, (local, remote, args) -> {
+        if (INSTALLING.get() && followerId.equals(local) && ((InstallSnapshotRequestProto) args[0]).hasNotification()) {
+          NOTIFICATIONS.incrementAndGet();
+        }
+        return false;
+      });
+
+      // Keep writing during the installation: each new entry wakes up the LogAppenders.
+      final AtomicBoolean stopWriting = new AtomicBoolean();
+      final Thread writer = new Thread(() -> {
+        try (RaftClient client = cluster.createClient(leader.getId())) {
+          for (int i = 0; !stopWriting.get(); i++) {
+            if (client.io().send(new RaftTestUtil.SimpleMessage("w" + i)).isSuccess() && INSTALLING.get()) {
+              CLIENT_WRITES.incrementAndGet();
+            }
+          }
+        } catch (Exception e) {
+          LOG.info("writer stopped", e);
+        }
+      }, "writer");
+      writer.setDaemon(true);
+
+      try {
+        writer.start();
+        // Restart the follower; it has to install the leader snapshot via notification.
+        cluster.restartServer(followerId, false);
+        // Only wait until the leader learns that the snapshot is installed: SimpleStateMachine4Testing
+        // cannot reload a snapshot on top of an existing state, so the follower is not checked for catching up.
+        JavaUtils.attempt(() -> {
+          Assertions.assertTrue(INSTALL_END_NANOS.get() > 0, "install not finished");
+          Assertions.assertTrue(RaftServerTestUtil.getLogAppenders(leader)
+              .filter(a -> a.getFollowerId().equals(followerId))
+              .anyMatch(a -> a.getFollower().getSnapshotIndex() >= leaderSnapshot.getIndex()), "leader not informed");
+        }, 60, ONE_SECOND, "snapshot installed", LOG);
+      } finally {
+        stopWriting.set(true);
+        writer.join(10_000);
+        CodeInjectionForTesting.remove(GrpcServicesImpl.GRPC_SEND_SERVER_REQUEST);
+        CodeInjectionForTesting.remove(INSTALL_SNAPSHOT_INJECTION);
+      }
+
+      final long installMs = TimeUnit.NANOSECONDS.toMillis(INSTALL_END_NANOS.get() - INSTALL_START_NANOS.get());
+      final long heartbeatMs = RaftServerConfigKeys.Rpc.timeoutMin(cluster.getProperties())
+          .toLong(TimeUnit.MILLISECONDS) / 2;
+      LOG.info("During the {}ms installation ({} client writes): {} notifications and {} heartbeats"
+          + " (heartbeat interval {}ms)", installMs, CLIENT_WRITES.get(), NOTIFICATIONS.get(), HEARTBEATS.get(),
+          heartbeatMs);
+
+      Assertions.assertTrue(installMs >= INSTALL_DELAY.toLong(TimeUnit.MILLISECONDS), "the installation was not slow");
+      Assertions.assertTrue(CLIENT_WRITES.get() > 0, "no client writes during the installation");
+      Assertions.assertEquals(0, APPEND_ENTRIES_WITH_ENTRIES.get(), "sent log entries during the installation");
+      // At most about one notification and one heartbeat per heartbeat interval (with 3x slack).
+      final long rpcs = NOTIFICATIONS.get() + HEARTBEATS.get();
+      final long maxRpcs = 3 * installMs / heartbeatMs;
+      Assertions.assertTrue(rpcs <= maxRpcs, () -> "Too many RPCs during the installation: " + NOTIFICATIONS.get()
+          + " notifications + " + HEARTBEATS.get() + " heartbeats = " + rpcs + " > " + maxRpcs);
+    }
+  }
+
+  private static void writeMessages(MiniRaftClusterWithGrpc cluster, RaftServer.Division leader, int n)
+      throws Exception {
+    try (RaftClient client = cluster.createClient(leader.getId())) {
+      for (int i = 0; i < n; i++) {
+        Assertions.assertTrue(client.io().send(new RaftTestUtil.SimpleMessage("m" + i)).isSuccess());
+      }
+    }
+  }
+}